There is ongoing work towards building FMI v2 support for SUMO by utilizing libsumo.
The initial goal is to build a prototype which is able to load and run a SUMO simulation via
libsumocpp while supporting a few selected libsumo functions.
Currently, these functions are (cf. ScalarVariables in
libsumo functions are currently added manually for the prototype. There are plans to automatically generate available libsumo functions for the FMI.
The current architecture to build the functionality for a Functional Mockup Unit (FMU) is as follows (corresponding source code is located in
Due to limitations of version 2 of the FMI standard, e.g., output scalar variables cannot be given parameters directly, we implemented a stateful approach.
For example, if we need the current position of a vehicle, we retrieve the (output) value of the
vehicle.getPosition scalar variable.
However, due to the above-mentioned limitation of FMI v2, we first need to set the input parameter value for
vehicle.getPosition, which is the vehicle ID.
For this purpose, we utilize the generic
setGetterParameters input scalar variable right before retrieving the output variable, as indicated by the following Python code snippet:
fmu.setString([valueRefs['setGetterParameters']], ["ego"]) resultList = fmu.getString([valueRefs['vehicle.getPosition']])
setGetterParameters, as its name implies, is a generic catch-all input scalar variable to set the input parameters for the subsequent retrieval of an output scalar variable's value.
In case of multiple input parameters, e.g., for
vehicle.getParameterWithKey, parameter values need to be concatenated to a single string with a space delimiter.
Similarly, output values are returned as a concatenated string with the same delimiter:
fmu.setString([valueRefs['setGetterParameters']], ["ego meaningOfLife"]) resultList = fmu.getString([valueRefs['vehicle.getParameterWithKey']]) key = resultList.decode('UTF-8').split() value = resultList.decode('UTF-8').split()
This convention of passing values as concatenated strings stems from the limitation of FMI v2 that vectors and other complex data types are not yet supported.
Interfacing the SUMO FMU with Python#
The FMPy Python package can be used to simulate as well as validate FMUs and is the recommended way to interface the SUMO FMU with Python. Installing FMPy on Linux is done with:
python -m pip install fmpy
An easy way to get started with FMPy is to check out the
runner.py for the SUMO fmi test suite located in
tests/complex/fmi (the code snippets from above can be found here, as well).